03_inspect_collaborator_panel

Niklas Rindtorff

1 Load packages

library(tidyverse)
library(pheatmap)
library(splots)
library(patchwork)
library(synergyfinder)
library(ggsignif)
library(broom)
library(ggrepel)
library(readxl)
library(platetools)

2 Load data

## read ctg data files
ctg_data <- lapply(list.files("~/tas/data/ctg_data/collab_plates/", 
    full.names = T, pattern = "1803"), function(f) read_tsv(f, 
    col_names = F, col_types = "cci") %>% `colnames<-`(c("name", 
    "well", "pcount")) %>% mutate(screen = f %>% str_split(pattern = "/") %>% 
    unlist %>% tail(1) %>% substr(., 1, nchar(.) - 4))) %>% bind_rows() %>% 
    separate(screen, c("date", "operator", "mithras", "experiment_id"), 
        sep = "_", remove = FALSE, extra = "merge") %>% # now I need to do some cleaning of badly named CTG datasets.
# Clara did this manually.
separate(experiment_id, c("organism", "id")) %>% # I now join the barcode annotation data with the photon
# counts
left_join(., read_excel("~/tas/data/barcodes/Organoid-Screening-Database-and-Barcodes.xlsx", 
    sheet = "Combinations") %>% drop_na() %>% rename(planned_date = `Screening Date`, 
    assay = `Image/CTG`, id = barcode) %>% filter(Library == 
    "L12")) %>% # I manually changed the name D007T01V036L11DM2 to
# D007T01V035L11DM2 to match the annotation table join
# compound annotation data
full_join(., lapply(list.files("~/tas/data/annotations/l12", 
    full.names = T, pattern = "csv"), function(f) read_csv(f)) %>% 
    bind_rows() %>% rename(well = destination.well, BlockID = drug_pair) %>% 
    select(concentration_pair, BlockID, well, substance) %>% 
    distinct() %>% separate(concentration_pair, c("Row", "Col"), 
    sep = "_") %>% separate(Row, c("Row"), extra = "drop") %>% 
    separate(Col, c("Col"), extra = "drop") %>% separate(BlockID, 
    c("DrugRow", "DrugCol"), sep = 1, remove = FALSE) %>% separate(well, 
    c("row", "col"), sep = 1, remove = FALSE) %>% mutate(combination = if_else(is.na(substance), 
    TRUE, FALSE))) %>% # now I perform some further conditional formatting first I
# have to change some old names
mutate(line = paste0(Donor, Tumor)) %>% rename(experiment = id) %>% 
    unite(id, c("date", "line", "Library", "experiment"), remove = FALSE) %>% 
    mutate(row_num = match(row, LETTERS[1:26]), col_num = as.numeric(col)) %>% 
    rename(drug = substance)

3 Inspect data

First I gain an overview of the collected data.

ctg_data %>% select(id, pcount) %>% arrange(id) %>% split(., 
    .$id) %>% lapply(., function(f) f$pcount %>% as.vector) %>% 
    plotScreen(., 2)

4 Correction

4.1 Spatial effects

Now I focus on column and row-wise effects. Row wise effects dominate clearly.

ctg_data %>% ggplot(aes(row_num, pcount)) + geom_point() + geom_smooth() + 
    facet_wrap(~id, nrow = 4) + theme_bw() + ggtitle("Spatial effects of rows")

## `geom_smooth()` using method = 'loess' and formula 'y ~ x'

# next figure
ctg_data %>% ggplot(aes(col_num, pcount)) + geom_point() + geom_smooth() + 
    facet_wrap(~id, nrow = 4) + theme_bw() + ggtitle("Spatial effects of columns")

## `geom_smooth()` using method = 'loess' and formula 'y ~ x'

I apply a conservative row-wise loess fit for each plate

## split data by plate and apply loess normalization
ctg_loess <- ctg_data %>% dplyr::select(row_num, col_num, pcount, 
    id) %>% # na values
drop_na() %>% # set the ctrl column 13 to NA mutate(pcount = ifelse(col_num
# == 13, NA, pcount)) %>%
split(.$id) %>% lapply(function(s) {
    ## loess fit. family is 'symmetric' to be robust to outliers
    fit <- loess(pcount ~ row_num + col_num, data = s, family = "symmetric")
    ## apply normalization
    tibble(norm_fac = fit$fitted) %>% cbind(s %>% drop_na(), 
        .) %>% mutate(pcount_norm = pcount - (norm_fac - median(norm_fac)))
}) %>% bind_rows() %>% full_join(ctg_data)

## Joining, by = c("row_num", "col_num", "pcount", "id")

Now I plot the results for rows and columns

ctg_loess %>% ggplot(aes(row_num, pcount_norm)) + geom_point() + 
    geom_smooth() + facet_wrap(~id, nrow = 4) + theme_bw() + 
    ggtitle("Spatial effects of rows")

## `geom_smooth()` using method = 'loess' and formula 'y ~ x'

# next figure
ctg_loess %>% ggplot(aes(col_num, pcount_norm)) + geom_point() + 
    geom_smooth() + facet_wrap(~id, nrow = 4) + theme_bw() + 
    ggtitle("Spatial effects of columns")

## `geom_smooth()` using method = 'loess' and formula 'y ~ x'

Again I plot the overview map for all plates. I think I have been able to reduce spatial effects.

ctg_loess %>% select(id, pcount_norm) %>% arrange(id) %>% split(., 
    .$id) %>% lapply(., function(f) f$pcount %>% as.vector) %>% 
    plotScreen(., 2)

5 QC

Now I look at the distribution of controls in the first screening replicate.

coi <- c("Staurosporine", "DMSO")
ctg_loess %>% mutate(line_date_exp = paste0(line, "_", date)) %>% 
    filter(drug %in% coi) %>% ggplot(aes(drug, pcount_norm)) + 
    # geom_violin(aes(group = drug)) + geom_jitter(alpha = 0.7) +
geom_boxplot(width = 1) + facet_grid(line ~ date) + theme_bw()

# ctg_loess %>% mutate(line_date_exp = paste0(line, '_',
# date)) %>% filter(drug %in% coi) %>% filter(combination ==
# TRUE) %>% ggplot(aes(id, pcount_norm, color = drug)) +
# #geom_violin(aes(group = drug)) + geom_boxplot(alpha = 0.7)
# + #facet_grid(line_date_exp ~ Library) + #facet_wrap(~line)
# + theme_classic()

I was able to calculate z-factors. They seem robust for all plates. However, line 27 shows poorer performance across both screening dates.

z_df <- ctg_loess %>% filter(drug %in% coi) %>% group_by(id, 
    drug, combination, Library, line, date, experiment) %>% summarise(sd = sd(pcount_norm, 
    na.rm = TRUE), mean = mean(pcount_norm, na.rm = TRUE)) %>% 
    group_by(id, combination, Library, line, date, experiment) %>% 
    summarise(zfactor = 1 - ((3 * sum(sd))/abs(range(mean)[1] - 
        range(mean)[2]))) %>% mutate(qc = if_else(zfactor < 0.25, 
    FALSE, TRUE))
z_df %>% mutate(query = substr(Library, 1, 5), query_conc = substr(Library, 
    6, 6)) %>% filter(combination != TRUE) %>% arrange(zfactor) %>% 
    ungroup() %>% mutate(id = factor(id, levels = id)) %>% ggplot(aes(id, 
    zfactor)) + geom_bar(stat = "identity") + theme_classic() + 
    scale_y_continuous(limits = c(0, 1)) + coord_flip()

# z_df_anno <- z_df %>% ungroup%>% mutate(id = paste0(id,
# '_', combination)) %>% dplyr::select(-experiment, -zfactor)
# %>% as.data.frame() %>% remove_rownames() %>%
# column_to_rownames('id') z_df %>% ungroup%>% mutate(id =
# paste0(id, '_', combination)) %>% dplyr::select(id,
# zfactor) %>% arrange(zfactor) %>% as.data.frame() %>%
# remove_rownames() %>% column_to_rownames('id') %>%
# pheatmap(., cluster_cols = FALSE, annotation_row =
# z_df_anno)

I prepare a publication style figure for the raw photon counts of uncombined controls

coi <- c("DMSO", "Staurosporine")
ctg_loess %>% filter(drug %in% coi & combination == FALSE) %>% 
    ggplot(aes(id, pcount_norm)) + geom_boxplot(aes(colour = drug), 
    width = 0.2, position = "identity") + theme_classic() + # facet_grid(line~date) + scale_y_continuous(limits =
# c(-1,250000)) + ggtitle('Before DMSO normalization') +
theme(axis.title.y = element_blank()) + scale_color_manual(values = rev(c("#999999", 
    "#000000"))) + ylab("Photon count") + # theme(axis.text.x = element_text(angle = 90, hjust = 1)) +
coord_flip() + ggsave("collab_ctrls_raw_byplate_uncombined.pdf", 
    width = 6, height = 4)

# cor(use = 'complete.obs') pheatmap()

Now I create a raw correlation plot for bowth replicates across all plates.

r <- ctg_loess %>% mutate(ctrl = ifelse(drug == "DMSO", "negative", 
    "none")) %>% mutate(ctrl = ifelse(drug == "Staurosporine", 
    "positive", ctrl)) %>% # filter(drug %in% coi) %>%
dplyr::select(line, Library, date, pcount_norm, well, ctrl) %>% 
    unite(id, c("line", "Library", "well")) %>% spread(date, 
    pcount_norm) %>% rename(rep1 = `180320`) %>% rename(rep2 = `180327`) %>% 
    dplyr::select(rep1, rep2) %>% as.matrix() %>% cor() %>% .["rep1", 
    "rep2"]
ctg_loess %>% # filter(col_num != 13) %>%
mutate(ctrl = ifelse(drug == "DMSO", "negative", "none")) %>% 
    mutate(ctrl = ifelse(drug == "Staurosporine", "positive", 
        ctrl)) %>% # filter(drug %in% coi) %>%
dplyr::select(line, Library, date, pcount_norm, well, ctrl) %>% 
    unite(id, c("line", "Library", "well")) %>% spread(date, 
    pcount_norm) %>% rename(rep1 = `180320`) %>% rename(rep2 = `180327`) %>% 
    ggplot(aes(rep1, rep2)) + geom_point(alpha = 0.2) + # geom_point(alpha = 0.2, aes(color = ctrl)) +
# scale_color_manual(values=c('#E69F00', '#999999',
# '#56B4E9')) +
geom_abline(intercept = 0, slope = 1) + theme_classic() + ggtitle(paste0("r (pearson) = ", 
    r %>% round(2))) + xlab("replicate 1") + ylab("replicate 2") + 
    scale_y_continuous(breaks = c(0, 250000, 5e+05, 750000)) + 
    coord_fixed() + ggsave("collab_pearson_raw.pdf", width = 4, 
    height = 4)

I plot the distributions of photon counts for both plates before normalization.

ctg_loess %>% mutate(line_date_exp = paste0(line, "_", date)) %>% 
    ggplot(aes(date, pcount_norm)) + geom_violin(aes(group = date, 
    color = date)) + # geom_jitter(alpha = 0.7) +
geom_boxplot(data = ctg_loess %>% filter(drug == "DMSO"), aes(group = date), 
    width = 0.25) + facet_grid(line ~ Library) + theme_bw() + 
    theme(axis.text.x = element_text(angle = 90, hjust = 1))

Now I normalize all photon counts to their respective DMSO controls

ctg_ln <- left_join(ctg_loess, ctg_loess %>% dplyr::select(id, 
    drug, pcount_norm) %>% filter(drug == "DMSO") %>% group_by(id) %>% 
    summarise(neg_control_median = median(pcount_norm, na.rm = TRUE))) %>% 
    mutate(viability = pcount_norm/neg_control_median)

## Joining, by = "id"

Moreover, I crop the viability to be always >= 0. Values of <0 are an artifact of loess smoothing.

ctg_ln <- ctg_ln %>% mutate(viability = ifelse(viability < 0, 
    0, viability))

Now I repeat my diagnostic plots.

ctg_ln %>% mutate(line_date_exp = paste0(line, "_", date)) %>% 
    ggplot(aes(date, viability)) + geom_violin(aes(group = date, 
    color = date)) + # geom_jitter(alpha = 0.7) +
geom_boxplot(data = ctg_ln %>% filter(drug == "DMSO"), aes(group = date), 
    width = 0.25) + facet_grid(line ~ Library) + theme_bw() + 
    theme(axis.text.x = element_text(angle = 90, hjust = 1))

r <- ctg_ln %>% mutate(ctrl = ifelse(drug == "DMSO", "negative", 
    "none")) %>% mutate(ctrl = ifelse(drug == "Staurosporine", 
    "positive", ctrl)) %>% # filter(drug %in% coi) %>%
dplyr::select(line, Library, date, viability, well, ctrl) %>% 
    unite(id, c("line", "Library", "well")) %>% spread(date, 
    viability) %>% rename(rep1 = `180320`) %>% rename(rep2 = `180327`) %>% 
    dplyr::select(rep1, rep2) %>% as.matrix() %>% cor() %>% .["rep1", 
    "rep2"]
ctg_ln %>% # filter(col_num != 13) %>%
mutate(ctrl = ifelse(drug == "DMSO", "negative", "none")) %>% 
    mutate(ctrl = ifelse(drug == "Staurosporine", "positive", 
        ctrl)) %>% # filter(drug %in% coi) %>%
dplyr::select(line, Library, date, viability, well, ctrl) %>% 
    unite(id, c("line", "Library", "well")) %>% spread(date, 
    viability) %>% rename(rep1 = `180320`) %>% rename(rep2 = `180327`) %>% 
    ggplot(aes(rep1, rep2)) + geom_point(alpha = 0.2) + # geom_point(alpha = 0.2, aes(color = ctrl)) +
# scale_color_manual(values=c('#E69F00', '#999999',
# '#56B4E9')) +
geom_abline(intercept = 0, slope = 1) + theme_classic() + ggtitle(paste0("r (pearson) = ", 
    r %>% round(2))) + xlab("replicate 1") + ylab("replicate 2") + 
    scale_y_continuous(breaks = c(0, 0.5, 1, 1.5)) + geom_hline(yintercept = 1) + 
    geom_vline(xintercept = 1) + coord_fixed() + ggsave("collab_pearson_norm.pdf", 
    width = 4, height = 4)

coi <- c("DMSO", "Staurosporine")
ctg_ln %>% filter(drug %in% coi & combination == FALSE) %>% ggplot(aes(id, 
    viability)) + geom_boxplot(aes(colour = drug), width = 0.2, 
    position = "identity") + theme_classic() + # facet_grid(line~date) + scale_y_continuous(limits =
# c(-1,250000)) + ggtitle('Before DMSO normalization') +
theme(axis.title.y = element_blank()) + scale_color_manual(values = rev(c("#999999", 
    "#000000"))) + ylab("Photon count") + # theme(axis.text.x = element_text(angle = 90, hjust = 1)) +
coord_flip() + ggsave("collab_ctrls_norm_byplate_uncombined.pdf", 
    width = 6, height = 4)

6 Calculate synergy

I consider using the “synergyfinder” package, published by He et al. at the FIMM. First I have to wrangle the data into a standardized input format. For this I write an ugly function.

coi <- c("Staurosporine", "DMSO")

syn_drug <- ctg_ln %>%
    filter(!drug %in% coi) %>%
    #filter(query != "L11DM") %>%
    #filter(combination == TRUE) %>%
    rename(rv = viability) %>%
    #setup the easy stuff
    dplyr::transmute(
      #BlockID = paste(query, drug, sep = "_"),
           Response = rv*100,
           Replicate = ifelse(date == "180320", 1, 2),
           DrugRow = DrugRow,
           DrugCol = DrugCol,
           #build matrix
           Col = Col,
           Row = Row,
           #define concentrations
           ConcRow = ifelse(Row == 6, 0, 25/(5^(as.numeric(Row)-1))), #assuming the initila concentration is 50
           ConcCol = ifelse(Col == 6, 0, 25/(5^(as.numeric(Col)-1))),
           ConcUnit = "uM",
      line = line, 
      BlockID = BlockID
      
           ) 
# %>% 
#   group_by(Replicate, Col, Row, line, BlockID, DrugRow, DrugCol, ConcRow, ConcCol, ConcUnit) %>% 
#   summarise(Response = mean(Response))
  
  # #Next I add the DMSO controls for the query drug concentrations
  # syn_drug <- full_join(syn_drug %>% drop_na(), syn_drug %>% filter(is.na(BlockID)) %>%
  #                         group_by(DrugRow, Row, Replicate, line, DrugCol, Col, ConcRow, ConcCol, ConcUnit) %>% 
  #                         summarise(Response = mean(Response, na.rm = TRUE))  %>% 
  #   do(cbind(., BlockID = paste(.$DrugRow, syn_drug %>% filter(DrugCol != "DMSO") %>% .$DrugCol %>% unique, sep = "_"))))
  # #Finally I add the last missing piece, the DMSO ~ DMSO control fields
  # 

Now I define a wrapper function that can apply multiple synergy calculations and report the results in a tidy format. For now I stick to calulating the average synergy scores.

tidy_synergy = function(df, me){ lapply(as.list(me), function(m, f = df){
  #hacky way of ignoring cases in which the synergy scores could not be calculated
  return(tryCatch(f %>% 
      ReshapeData(., data.type = "viability") %>%
      #PlotDoseResponse() %>%
      CalculateSynergy(method = m, correction = TRUE) %>%
    tibble(BlockID = .$drug.pairs %>% simplify() %>% .[4],
           method = .$method,
           score_average = .$scores %>% .[[1]] %>% .[-1,-1] %>% simplify() %>% mean(),
           score_sd = .$scores %>% .[[1]] %>% .[-1,-1] %>% simplify() %>% sd()) %>%
    select(BlockID, method, score_average, score_sd) %>% distinct(), 
    error=function(e) NULL))}
) %>% bind_rows()
  #PlotSynergy(type = "all")
  }

I apply the function on the dataset and save/load it.

collab_syn <- syn_drug %>% group_by(line, Replicate, BlockID) %>% 
    do(tidy_synergy(df = ., me = c("ZIP", "HSA", "Bliss", "Loewe"))) %>% 
    ungroup() %>% # complete the df for scores that could not be calculated
mutate(method = as.factor(method), line = as.factor(line)) %>% 
    complete(BlockID, method, Replicate, line) %>% mutate(score_average = ifelse(is.nan(score_average), 
    NA, score_average))
save(collab_syn, file = "collab_syn_complete.Rdata")

7 Plot data

7.1 Plot synergy

As an example I plot a complete treatment surface for one treatmetn of Trametinib + PRI

library(plotly)
boi = c("AB")
tmp <- syn_drug %>% filter(line == "D007T01", BlockID == boi, 
    Replicate == 1) %>% group_by(Replicate, Col, Row, line, BlockID, 
    DrugRow, DrugCol, ConcRow, ConcCol, ConcUnit) %>% summarise(Response = mean(Response)) %>% 
    ungroup %>% dplyr::select(Col, Row, Response) %>% spread(Row, 
    Response) %>% as.matrix()
class(tmp) <- "numeric"
p <- plot_ly(z = ~tmp) %>% add_surface()
p

I check how often for which synergy score no result could be calculated

load("collab_syn_complete.Rdata")
collab_syn %>% group_by(method) %>% summarise(n = is.na(score_average) %>% 
    sum())

I plot the calculated synergy scores

collab_syn %>% mutate(Replicate = as.factor(Replicate)) %>% ggplot(aes(BlockID, 
    score_average)) + geom_point(aes(color = Replicate, shape = line), 
    size = 3) + theme_classic() + scale_color_manual(values = c("#4885ed", 
    "#db3236")) + facet_wrap(~method) + geom_hline(yintercept = 0)

## Warning: Removed 17 rows containing missing values (geom_point).

7.2 Plot treatment data

Now I plot heatmaps for all treatments. The heatmaps are in this order:

7.2.1 Heatmaps

syn_drug %>% group_by(Replicate, Col, Row, line, BlockID, DrugRow, 
    DrugCol, ConcRow, ConcCol, ConcUnit) %>% summarise(Response = mean(Response)) %>% 
    group_by(line, BlockID, Replicate) %>% do(ungroup(.) %>% 
    dplyr::select(Col, Row, Response) %>% spread(Row, Response) %>% 
    .[, -1]) %>% nest(-c(BlockID, line, Replicate)) %>% arrange(BlockID, 
    line, Replicate) %>% dplyr::select(-data)

syn_drug %>% group_by(Replicate, Col, Row, line, BlockID, DrugRow, 
    DrugCol, ConcRow, ConcCol, ConcUnit) %>% summarise(Response = mean(Response)) %>% 
    group_by(line, BlockID, Replicate) %>% do(ungroup(.) %>% 
    dplyr::select(Col, Row, Response) %>% spread(Row, Response) %>% 
    .[, -1]) %>% nest(-c(BlockID, line, Replicate)) %>% arrange(BlockID, 
    line, Replicate) %>% mutate(data = map(data, ~as.matrix(.x))) %>% 
    pull(data) %>% walk(~pheatmap(.x, cluster_rows = FALSE, cluster_cols = FALSE))

7.2.2 Dose-response curves

Now I plot the dose-response curves for all treatment seperately

library(drc)
library(forcats)
tidy_drc_fit <- function(df){  
  # predictions and confidence intervals.
  org.fits <- expand.grid(conc=exp(seq(log(min(df$conc)), log(max(df$conc)), length=1000)))
  
  fits <- drc::drm(data = df, pc_norm ~ conc, fct=LL.4(), na.action = na.omit) %>%
  predict(., newdata = org.fits, interval = "confidence")
  
  org.fits %>%
    mutate(p = fits[,1],
           pmin = fits[,2],
           pmax = fits[,3],
           line = df$line %>% unique(),
           drug = df$drug %>% unique()) %>%
    full_join(df) %>%
  return(.)
}


coi <- c("C")

fit_syn_drug <- syn_drug %>% 
  mutate(conc = Row,
         pc_norm = Response,
         drug = DrugRow) %>%
  filter(Col == 6) %>%
  dplyr::select(line, drug, pc_norm, conc) %>%
  #filter(line %in% loi, drug %in% coi) %>% 
  mutate(conc = as.numeric(conc)) %>%
  group_by(drug, line) %>%
  do(., tidy_drc_fit(.)) %>%
  rbind(., syn_drug %>% 
  mutate(conc = Col,
         pc_norm = Response,
         drug = DrugCol) %>%
  filter(Row == 6) %>%
  dplyr::select(line, drug, pc_norm, conc) %>%
  #filter(line %in% loi, drug %in% coi) %>% 
  mutate(conc = as.numeric(conc)) %>%
  group_by(drug, line) %>%
  do(., tidy_drc_fit(.))) %>% 
  ungroup() %>%
  filter(conc <= 5) %>% # removing DMSO fit
  mutate(drug = if_else(drug == "A", "Trametinib",
                        if_else(drug == "B", "PRI-724",
                                if_else(drug == "C", "UNC-0638", 
                                        if_else(drug == "D", "Vorinostat", 
                                                if_else(drug == "E", "Rosiglitazone",
                                                        if_else(drug == "F", "5-FU", "no match"))))))) %>% 
  mutate(drug = factor(drug, levels = c("Trametinib", "PRI-724", "UNC-0638", "Vorinostat", "Rosiglitazone", "5-FU")))

coi <- c("Trametinib", "PRI-724")
fit_syn_drug %>% filter(drug %in% coi) %>% ggplot() + geom_point(aes(conc, 
    pc_norm), color = "#ff9900", alpha = 0.4) + geom_line(aes(conc, 
    p), color = "#146eb4", size = 1.2) + theme_bw() + scale_y_continuous(limits = c(0, 
    120)) + # geom_smooth(formula = y ~ x) +
geom_hline(yintercept = 100) + # geom_hline(yintercept = 0) +
facet_grid(drug ~ line) + scale_x_reverse() + ylab("Relative Viability") + 
    xlab("Concentration") + ggsave("AB_dose_respone.pdf", width = 4, 
    height = 3)

## Warning: Removed 3590 rows containing missing values (geom_point).

## Warning: Removed 6 rows containing missing values (geom_path).

## Warning: Removed 3590 rows containing missing values (geom_point).

## Warning: Removed 6 rows containing missing values (geom_path).

coi <- c("UNC-0638", "Vorinostat")
fit_syn_drug %>% filter(drug %in% coi) %>% ggplot() + geom_point(aes(conc, 
    pc_norm), color = "#ff9900", alpha = 0.4) + geom_line(aes(conc, 
    p), color = "#146eb4", size = 1.2) + theme_bw() + scale_y_continuous(limits = c(0, 
    120)) + # geom_smooth(formula = y ~ x) +
geom_hline(yintercept = 100) + # geom_hline(yintercept = 0) +
facet_grid(drug ~ line) + scale_x_reverse() + ylab("Relative Viability") + 
    xlab("Concentration") + ggsave("CD_dose_respone.pdf", width = 4, 
    height = 3)

## Warning: Removed 3589 rows containing missing values (geom_point).

## Warning: Removed 6 rows containing missing values (geom_path).

## Warning: Removed 3589 rows containing missing values (geom_point).

## Warning: Removed 6 rows containing missing values (geom_path).

coi <- c("Rosiglitazone", "5-FU")
fit_syn_drug %>% filter(drug %in% coi) %>% ggplot() + geom_point(aes(conc, 
    pc_norm), color = "#ff9900", alpha = 0.4) + geom_line(aes(conc, 
    p), color = "#146eb4", size = 1.2) + theme_bw() + scale_y_continuous(limits = c(0, 
    120)) + # geom_smooth(formula = y ~ x) +
geom_hline(yintercept = 100) + # geom_hline(yintercept = 0) +
facet_grid(drug ~ line) + scale_x_reverse() + ylab("Relative Viability") + 
    xlab("Concentration") + ggsave("EF_dose_respone.pdf", width = 4, 
    height = 3)

## Warning: Removed 3590 rows containing missing values (geom_point).

## Warning: Removed 6 rows containing missing values (geom_path).

## Warning: Removed 3590 rows containing missing values (geom_point).

## Warning: Removed 6 rows containing missing values (geom_path).

8 Abadoned approaches

I list steps that are now covered by B. Rasucher or other alternative approaches in the .Rmd below.

knitr::knit_exit()